XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL

Pesquisa Operacional na Gestão da Segurança Pública

16 a 19
Setembro de 2014
Salvador/BA

A PRINCIPAL-AGENT APPROACH FOR THE CROWDING-OUT EFFECT IN

INFORMATION TECHNOLOGY OUTSOURCING CONTRACTS UNDER ASYMMETRIC
INFORMATION

Thyago Celso Cavalcante Nepomuceno

Universidade Federal de Pernambuco
Av. Prof Moraes Rego, 1235 – Cidade Universitária, Recife – PE
thyago.nepomuceno@yahoo.com
,
Ana Paula Cabral S. Costa
Universidade Federal de Pernambuco
Av. Prof Moraes Rego, 1235 – Cidade Universitária, Recife – PE
apcabral@ufpe.br

ABSTRACT

It has been claimed in the literature the role of punishment as promoting a behavior that can align
both objectives the client and the vendor into a set of actions desirable for both. Penalty clauses are
commonly used in outsourcing contracts to guarantee a certain level of quality for the good or service or
to assure that the delivery date will be respected. However, some grades of penalties instead of avoid the
behavior of the vendor from the undesirable by the client works as an incentive to the commitment of
more irregularities. The current study derives a general cost structure from the software engineering
literature for a typical software vendor and analyzes through a principal-agent model when the imposition
of a fine by the client (the principal) whenever the vendor (the agent) delays the delivery of a software
can stimulate the vendor to delay more, a behavior named crowding-out effect.

KEYWORDS: Incentive Theory; Information Technology Outsourcing; Software Engineering;

Crowding-out Effect

Main area: TEL&SI – Operational Research in Telecommunications and Information Systems, OA

– Other application in Operational Research, SE – Operational Research in Services.

3031
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

1. Introduction

Information Technology Outsourcing (ITO) constitute a typical principal-agent problem in game

theory analysis, where a client, firm or a hirer (the principal) wants to maximize their profit by delegate
one or a set of activities regarded to the information technology of the business for another person or
organization, a vendor or provider (the agent) who will produce what is demanded and charge a certain
payment previously signed in a contract in return for the task done (Jensen and Meckling, 1976;
Eisenhardt, 1989; Whang, 1992; Keil, 2005; Dey et al. 2010; Singer and Donoso, 2011). The transference
of responsibility of create or supply the technological structure that will be critical to the business enables
the principal to take more efforts into the core and more profitable activity for the company (Willcocks
and Feeny, 2006; Lacity et al., 2009), reduces uncertainty and provides costs saving (Lacity and
Willcocks, 1998; Watjatraku, 2005), enables resources for other activities and reduces the amount of risk
with the acquirement of ability and expertise by the agent who holds the benefits of the economies of
scale and scope (Banker, 1989; Cachon and Haker, 2002; Bustinza et al., 2010).
There are situations which, rather than following the conventional thought, penalty clause in a
typical outsourcing contract decreases the agent's intrinsic motivation to perform a task, providing
incentives for the principal to commit even more the prohibited behavior. This situation has been called
crowding-out effect by Frey and Jegen (2001), also known as the overjustification effect by the
psychological theory of self-perception (Deci, 1971), implying that motivations may be undermine when
no-monetary interaction becomes an explicit monetary one. The evidence for this effect when involves
penalty is mentioned by Gneezy and Rustichini (2000) who analyzed the imposition of a monetary fine
for late-coming parents in a group of daycare centers in Israel. The fine, thought to induce the parents’
behavior into a reduction of the late, resulted in an opposite effect increasing the number of late-coming
by the parents. Also, Gneezy (2003) applied a proposer-responder game to demonstrate that the proposer
gives on average more money to the responder in pure situations (those with no reward or penalty relative
to the amount given), than in moments when the responder can penance a fine over the value the proposer
decides to give.
Despite the relevance of these evidences a model for this effect of penalty in contracts has never
been formalized mathematically and the literature do not draw any conclusions whether this effect is
visible in a contractual relationship between companies with different activities as it is an information
technology outsourcing scenario. In fact, such an effect has been not receiving any attention in the
discussion of the formulation of outsourcing information technology contracts. A pioneering model
analysis and a posterior empirical research may add a fundamental value for the discussion when
incentives may or may not work in a given type of contractual relationship, moreover, knowing what kind
of incentives vendors and clients are subject may grant a great strategic advantage to the possessor of this
information both for the formulation of their beliefs and actions, as to strategically influence the beliefs
and actions of the other part. Also, the model we are about to draw may allow the decision making analyst
to understand the best way to conduct the scenario, can minimize uncertainty and optimize the accuracy
of the forecasts into an outcome and possible equilibrium in the game.

2. First Time Contract Model

According to the conventional software engineering literature (Boehm, 1981; Londeix, 1987;
Boehm et al., 2000; Sommerville, 2011) the vendor's cost structure can be expressed as an exponential
function of the overall effort provided. The cost structure for the agent do not usually increase linearly
with software type (where it represents a high or low quality) due mainly to the side effects observed in
the project testing phase: When changes in specifications or errors resolutions leads to another unexpected
problems and errors, side effects of the original one. The result is that the more quality the software
contract specifies, the more complex the project will be and more anomalies may occur during the
development code phase. The sequential repeated game context infers that the agent is also concerning
about her reputation and does not focus her strategy only on the currently payment, but also on the long

3032
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

term profit that later outsourcing contracts and re-contracts dealt with the same client may guarantee. The
problem the agent faces is summarized by a choice of T, represented by:
ொ್
‫ܷݔܽܯ‬௔ ൌ ܲ ൅ ܲ௙ ൅ ܶ െ ‫ܥ‬௙ െ ‫ ˦ܥ‬௉ ‫ܭܮ‬ (1)
் ೑்

Where Cє is the vendor’s cost sensitivity to the effort assigned by ( QbLK)/ܲ௙ T. The coefficient T
is a function of the delivery date assigned in the contract (T1) and the effective date the agent delivers the
product (T2) so that T = T2 – T1. A high T means that the agent took more time to deliver the software
product, or delayed its delivery (high negative T2) and a low T (even negative) has the meaning that the
agent delivered the software product in less time (low T2), since the delivery date (T1) is fixed. The
fraction Qb/ܲ௙ T is the ratio between the quality of the software system Q (observed by the vendor through
the contract type) and the expected payment ܲ௙ weighted by the time the agent spends to develop the
software system T. As the vendor notes the quality of the software is greater than the expected gain they
might have, the ratio adds to the overall effort cost (Q/ܲ௙ ܶ > 1), otherwise, when the vendor believes the
expected payment will be greater than the endeavor of settle that type of software quality, the ratio
reduces the effort cost (Q/PT < 1), and when both contributes similarly to the perception of the agent,
only the technical parameters LK will matter for the effort-related cost (Q/PT = 1).
The exponent b clarifies the non-linear relationship previously described by the side effects of a
development code and test phase, and it is usually estimated between 1 and 1,5 (Boehm et al., 2000;
Sommerville, 2011). The exponent is sensitive to the novelty and risk of the project, it can be influenced
from others parameters such as the bad performance of the development team in understand or use an
existing code, and the fact b ≥ 1 has the interpretation that positive changes in the product quality causes
positive variations in the overall effort cost (∂C(e)/∂Q > 0) and that variation in the effort cost is
increasing (∂²C(e)/∂Q² > 0). The parameters LK are constants that capture the developer's lack of
experience, skills or the immaturity level of the development team, the unproductivity of the labor which
we assign by L, and the lack of capabilities of the software tools together with the nonexistence of a reuse
code, the capital's unproductivity which we assign by K. The agent still holds a fixed cost Cf over which
must bear even without producing anything. The gain the vendor ensures with the outsourcing contract is
composed by the payment the current project will guarantee, P, plus an expected profit counted from
forthcoming contracts the client might outsource to this vendor, Pf.
The principal derives utility from the quality of the software system so that a contract that
specifies a high quality software i. e. with many features or more comfortable interaction for the user is
always preferable for the client than a contract which results a low quality product. The utility of the
principal decreases as the agent attempts to delay the delivery of the software, but not in the same amount.
A High Cost Client (HCC) is assumed to have more difficulties to work without the technology product
of the outsourcing contract than a Low Cost Client (LCC). In other words, it is more costly for the High
Cost Client to deal the same amount of business in the lacking of the software system requested than it is
for the Low Cost Client, and for that reason, the High Cost Client values a shorter delivery time (with no
delays) and a high quality software with more intensity than the Low Cost Client. Thus the problem the
client faces is summarized by a choice of Q, once she does not have any power over the action of the
agent into to delay or not delay, and it could be represented by:

‫ܷݔܽܯ‬௣ ൌ ‫ݑ‬ଵ ܳ ൅ ‫ݑ‬ଶ ‫ܩ‬௙ െ ‫ݑ‬ଷ ܶ െ ܲ െ ܲ௙ (2)

ொ

Where u1 means how much the principal values a software quality type. A big u1 for a high
quality Q (or a small u1 for a low Q) tell us that the client cares too much about the software product
specifications and quality, what makes us assume we are probably dealing with a High Cost Client,
whereas a small u1 for a high Q (or a big u1 for a low Q) make us induce that type of principal as a Low
Cost Client since the client values fewer the software product quality. The parameter G is the expected

3033
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

gain the client might have with future outsourcing contracts and it increases with the current performance
of the agent, i. e., the more quality the software system has in less time of development, more profitable it
will be for the client now, and greater it is for the client the probability of been outsourcing the
technological issues and infrastructure for a high productivity vendor, and higher is the future profit. The
coefficient u2 captures the level of importance that the principal gives for those expected profits whereas
u3 is the sensitivity of the client unto the delay of the project conclusion. It has the same meaning from the
u1 coefficient described above: A big u3 for a high time of development T (or a small u3 for a low T) tell
us that the principal cares too much about the software product delivery time, what makes us assume we
are probably dealing with a High Cost Client, whereas a small u3 for a high T (or a big u3 for a low T)
make us induce that type of principal as a Low Cost Client since the valuation is small for software
product time to release. In order to interpret the performance of the outsourcing contract we must look for
the total social profit, obtained when we join the best response from both sides into a First Best Solution
(Dey et al. 2010). Summing (1) with (2), the problem then turn into:

ொ್
‫ ܷ ݔܽܯ‬ൌ ‫ݑ‬ଵ ܳ ൅ ‫ݑ‬ଶ ‫ܩ‬௙ ൅ ܶ െ ‫ݑ‬ଷ ܶ െ ‫ܥ‬௙ െ ‫˦ܥ‬ ‫ܭܮ‬ (3)
ொǡ் ௉೑ ்

The First-Order Conditions (FOC) for the problem above are:

భ
డ௎ ஼˦ ொ ್ ௅௄ ஼˦ ொ ್ ௅௄ మ

డ்
ൌ ͳ െ ‫ݑ‬ଷ െ ௉೑ ் మ
ൌ Ͳ ֜ ܶ ൌ ൬௉ ൰ (4)
೑ ି௨య ௉೑
భ
డ௎ ஼˦ ொ ሺ್షభሻ ௅௄ ௉ ்௨భ ሺ್షభሻ
డொ
ൌ ‫ݑ‬ଵ െ ܾ ௉೑ ்
ൌͲ֜ܳൌ ቀ ೑ ቁ (5)
௕஼˦ ௅௄

Those are the optimal choice for both the agent and the principal a priori. The First Best
Conditions derived above gives us some expected but not uninteresting results. At first, equation (4) tell
us that the best choice of T for the agent is positively related to her limitations on capital and labor (K and
L), the quality of the software product (Q) and the sensitivity to the variable cost related to the overall
effort (Cє), what makes perfect sense, once a high quality software, a large gap in productivity or a high
cost structure will persuade the agent to take more time to develop and hence delay the deliver. It is also
negatively related to expected payment the agent might have with later contracts (Pf), and this concern
adds to the choice of T more than itself because it is summed with itself weighted by a grade of
importance the principal considers an increase (or decrease) in the time of development (u3 Pf).
About the best quality choice the Equation (5) predicts that a change in the perception the
principal has over the quality of the software (u1) will affect positively the quality that will be demanded
in contract. In other words, if the client cares more about the quality, higher quality software will be
ordered. Also, if for some reason the client perceives that the vendor will have much to gain in future
contracts (Pf) this will give a great bargain power to the principal who will increase the quality of the
current one. We also see a positive relationship between the quality of the product and the time of
development (T) because, a priori, a greater number of days to perform a service implies in the possibility
of obtain a higher quality for the task than it would be if had a little time to do it. Thus, the longer the
delay, the higher the quality demanded by the client at first. On the other hand, when the principal feels
the agent will have much difficulties to develop the software ordered due to the limitation the vendor has
in relation to his capital, techniques or workers (K and L) or due to the high cost of the agent (Cє), she
will not be able to order a high quality software system, choosing a low one instead (inversely
relationship).
The difficult is increased by the parameter influenced by the novelty and risk of the software (b).
We may notice that the client has a dominant strategy since his utility increases with the quality of the
product, it will always be better to choose a high quality software than a low one, and since the agent has
no power of treat. We use backward induction to verify the best response of the agent to the optimum

3034
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

choice of quality by the principal. Substituting (5) in (4) and assigning b = 1,5 (which is very close to the
real estimations being in software engineering literature) we will have after all simplification process:

௕య ሺ஼˦ ௅௄ሻమ ି௨య ௕య ሺ஼˦ ௅௄ሻమ

ܶൌሾ య ሿ (6)
ቆ௉೑ ௨భమ ቇ

Equation (6) reveals some important concepts concerning the behavior of the agent in a sequential
game when considering, within the decision making, the optimal quality type of software chosen by the
principal. First we observe that the best answer as a choice of development time T is negatively related to
the expected gains the agent might have in the future and also to the degree of importance that the
principal gives about the software quality u1 This is not a surprising since the expected gains for future
contracts grows with a good reputation now, so the best choice to increase these gains would be not to
delay the delivery date, and, if it is possible, to further reduce the development time T to satisfy the
expected growth in these gains. Likewise, when the utility of the principal increases for a given software
the rational choice would be to provide the same software in the shortest time possible in order to keep
satisfied the utility of the client and also increase the probability of doing business again (and so increase
the expected gains).
About the upper side of the ratio, we have another logical proposition: the time in which the agent
takes into development depends positively on their productivity needs and technical constraints, i. e. the
deficiency expressed by the unproductivity of the workers, L, their equipment and development tools
expressed by K and the variable effort cost of producing a given software, Cє, weighted by the risk that a
new project can bring, b. As long as the deficiency increases, the vendor will take a longer time to
conclude and deliver the product software.
The more important consideration is the emergence of the parameter u3 in the equation. Since
CєKL are constants known by the agent representing the cost of producing that particular software and u1
and P are known approximately when the principal chooses the type of quality of the software, we have
the decision of the agent focused primarily on comparing the effect of parameter u3, which represents the
sensitivity of the client to changes in the development and delivery time of the software. The u3 parameter
reflects the intensity which the principal evaluates an anticipation in the date of delivery or a delay in the
same date. For this parameter we assume 3 different situations that could interpret each type of principal
we would be dealing with: If the parameter is very high (with value greater than 1), we know that the time
for the principal is valuable enough and we can deduct we are dealing with a High Cost Client. If this
parameter is too low (less than 1) we can deduct be dealing with a Low Cost Client, since this type of
principal appreciates fewer the time the agent has probability because this client is highly productive and
perform their tasks with greater ease and speed. Even for this type of principal (Low Cost Client) we have
the situation in which this parameter can be equal to 1, which means that the Principal evaluates a delay
(or advance) in the delivery date exactly equal to the value the agent has for her time. We discuss each of
the three frames pictured above about the parameter u3 and interpret its effect on the optimal choice of
development time by the agent.
First we analyze the most simple situation, when the agent believes that the principal values
highly the software development and delivery time (when u3 > 1), which would imply a loss for the agent
if she wishes to delay the delivery of the product, since the principal would not renew business with her.
When u3 > 1 we see that (bC1LK)² - (u3 bCєLK)² < 0 for all nonzero positive u3 Pf, which results in a
delivery time T < 0, meaning that the value of the choice of the variation in the software delivery time for
the agent T = (T2 – T1) is in fact negative, in other words, the time that the agent effectively deliver
product (T2) should be shorter than the time signed in the contract (T1) so that T is negative. Thus, if the
agent believes that the principal values highly her time, then the agent's choice would be even anticipate
the delivery date rather than delay it.

3035
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

Assuming now the principal evaluation of time is supposed to be the same as the evaluation the
agent has on her own time, that is, u3(T1 - T2) = (T1 - T2) resulting in u3 = 1. This hypothesis can be quite
feasible in our environment of incomplete or asymmetric information where one party knows nothing or
almost nothing about the other party. When the agent estimates the principal has the same priority on the
time as the agent has for hers the best response by the agent is not to delay the delivery of the software
since the closer the estimates are to each other, closer to zero the upper side of the equation will approach
and closer to zero is the change in T will be.
Finally, we observe the effects on the choice of T when u3 < 1, e.g, when the agent believes that
the principal does not matter very much about their time. From this point of view, the agent has no reason
to believe that the principal is a Low Cost Client since the only information available to the agent is the
quality of software, which it will always be high given the dominant strategy by the client. Thus, from
one way or the other, in a sequential game of asymmetric information which the principal may understand
about the agent’s framework but not the otherwise, the agent has an intrinsic incentive to never delay a
delivery of the product. The only reasonable hypothesis for an opposite behavior is when the agent has
strategic information provided by the client or obtained in some other way that justifies the fact that both
the principal valuation for the delay is small and the gains with futures contracts will not be diminished
by the delay. This hypothesis shatters the scenario in which we find ourselves, and should be taken into
consideration in the next section.

3. Penalty Contract Model

Let us now assume that the client afraid of a possible delay by the vendor (but not their
incentives) wishes to impose a fine for each day the agent delays the delivery of the software system,
counted from the date originally agreed. The model we are working with can now be represented as
follows:
ܳ௕
‫ ܷ ݔܽܯ‬ൌ ‫ݑ‬ଵ ܳ ൅ ‫ݑ‬ଶ ‫ܩ‬௙ െ ‫ݑ‬ଷ ܶ ൅ ݂ሺܶሻ െ ܲ െ ܲ௙ ൅ ቈܲ ൅ ܲ௙ ൅ ܶ െ ‫ݑ‬ସ ݂ሺܶሻ െ ‫ܥ‬௙ െ ‫˦ܥ‬ ‫ܭܮ‬቉
ொǡ் ܲ௙ ܶ

Where the parameter f represents the fine that the client requires and must be paid by the vendor
and u4 means the vendor's sensitivity to the cost of the penalty. The first expression (outside the brackets)
represents the profit to the principal while the second (inside the brackets) represents the profit to the
agent. It was put the two expressions together into one so that we may seek the best portrayal for the total
performance of this type of outsourcing contract ruled by clauses of penalties. The price of the fine has a
constant value equal to f for the principal, the same cannot be said for the evaluation the agent makes to
the same penalty. Given the sensitivity of the vendor to the price for the delay u4, it is reasonable to
assume that the degree of importance for the agent increases the lower the endowment of the firm and the
greater their costs and the price of the fine. The optimal choice for both the agent and the principal in this
contract can be found by the same way we did before. We use backward induction to verify the best
response of the agent to the optimum choice of quality by the principal. Once the client signals to the
agent her truthful evaluation through the imposition of a fine, there is no reason for us to keep thinking
that this signaling would not update the strategy of the agent. A high penalty fine indicates that the
principal highly values the time and a low penalty fine signals just the opposite. So we may assume the
agent's best estimative of the cost of the time for the principal is the amount of the fine imposed, and the
best choice of time by the agent is updated by this amount. Then, making u3 = f, and b = 1,5 we will have
after the simplifications:

௕య ሺ஼˦ ௅௄ሻమ ି௨ర ௙ሾ௕య ሺ஼˦ ௅௄ሻమ ሿ

ܶൌ య (7)
ቆ௉೑మ ௨భమ ቇ

3036
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

What equation (7) tell us is that the optimal choice for the vendor concerning the delivery time of
the software is solely based on a comparison of their costs in relation to the fine imposed, everything else
held constant. If the client imposes a fine f too high, and the vendor highly values the money they would
have to spend as payment of such fine (a high u4) then the equation turns to be negative (- u4 f [b³(Cє L
K)²] > b³(C1 L K )²) and T < 0, meaning that it is likely the vendor will anticipate rather than delay the
delivery date. Otherwise, if the client imposes a fine too low, and the vendor appreciates fewer the
amount of the money spent as payment of such fine (a low u4) then the equation turns to be positive (- u4 f
[b³(C1 L K)²] < b³(C1 L K )²) and T > 0, meaning that the vendor would delay the delivery date (occurrence
of the crowding-out effect). Any variation between these two results will depend solely on the comparison
made between the cost structure of the agent and how much that fine would harm the agent or support her.
Because now the agent knows for sure what type of principal she is negotiating, if the principal
imposes a lower penalty fine (in the perception of the agent), the agent may delay and still maintain an
expectation of future earnings from future outsourcing contracts with the same client. We have found an
interesting approach of the consequence that the crowding-out effect may outcome over an asymmetric
information sequential game: The fact that to impose a penalty over some irregularity such as a delay, as
we have analyzed, may cause further emergence of such irregularity instead of make it disappear. In the
outsource contract scenario it leads to a distortion of the previously observed balance of the game which
the client chose a high-quality software as the dominant strategy and the vendor responded by not
delaying as the dominant strategy. When the client attempts to guarantee that the vendor will not delay the
delivery imposing a fine and thinking that, if before the fine the vendor was not likely to delay, now with
the imposition of the fine the vendor will not delay for sure, this conclusion does not take into account the
incentives arising from the information asymmetry in this kind of game. The client thoughts of to depart
the chances of delay by the vendor is instead giving the agent the opportunity to delay by signaling her
cost structure to the vendor, who now updates her optimal strategy which may be to delay the delivery
date of the software product, when sooner she would not.

4. Conclusions

The model we just built deduced that an outsourcing of information technology characterized as a
sequential game of incomplete information, with potential future gains for both parties, culminates in the
principal, i. e. the client, having a dominant strategy which is to choose a contract of high quality
software, since the agent, i. e. the vendor, would always strive to ensure that quality for the client in order
to maintain a good reputation to future contracts and because the absence of adequate short-term
incentives to delay the date of delivery of the software. Given the uncertainty of the vendor to know what
kind of client they are dealing with (whether high or low cost), the expected payoff made by the current
payment plus a future value that are expected according current performance of the vendor will be higher
when the agent do not delay the delivery of the software product. So, unless the agent has strategic
information provided by the client or obtained in some other way that justifies the fact that both the
principal has a lower value for her time and the gains to the agent from futures contracts would not be
diminished with a delay, the vendor will never delay the software delivery in this sequential game. If the
agent has such information about the type of principal, outsourcing will no longer be characterized as a
sequential game of incomplete information, and the agent delay strategy may be drastically modified by
the reason of it being no longer dealing with expected payoffs, but now with certain payoffs.
We have demonstrate that the imposition of a fine works as a sign to the agent about the cost
structure of the principal and thereby her type (whether high or low cost), which can often result in a
crowding-out effect on the incentive structures where this grade of penalty may increase the delay, rather
than eliminate it, because now the agent updates her estimations about the type of the client according to
the price of the fine and the value of the fine for herself (that not necessarily must be equal to the price of
the fine for the client). If the fine is too high in the perception of the client but is not that expensive for the
vendor, or if both has a low valuation about the price of the fine, the agent would delay the delivery,
otherwise, if this fine is too high in the perception of both (but not high enough to be considered as unfair

3037
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

and breaks the contract) or if it is expensive for the vendor but not that much for the client, the vendor
agent would not delay the delivery of the software product. The conclusion our formal model derives
comes from the same conclusion Greiff et al. (2013) took about the strategic value the information
asymmetry can generate. In their study the authors conclude that the price that consumers are willing to
pay for a good or service on a Pay What You Want pricing scenario depends on how much information
these consumers possess regarding the costs of producers. If the supplier of a good or service has low
costs, the best strategy for her is to not reveal any anchor price so that consumers have no idea how much
they must pay. If the cost structure of the producer is high, this producer will raise more money revealing
a reference price for the consumers. The same analogy could form our independent outsourcing scenario:
If the client observes her nature as a Low Cost Client, the best optimum choice would be to choose a
contract with no penalty clauses in order to avoid a possible delay by the vendor. If the client is a High
Cost Client, the imposition of a fine may be the best strategy. The same analysis can be made into any of
these factors: a monetary value, number of days of delay or quality of the product and the conclusion shall
not be affected. However, an empirical approach to this effect in real IT outsourcing contracts and
services outsourcing contracts should add sufficient value to the theoretical gap our model leaves open.
Camlas Consulting, a firm specialized in management and contracts review for IT outsourcing in Europe,
found IT services contracts where it was cheaper for the supplier to pay a fine than putting time, effort
and money to meet the agreed service levels. In fact, a more practical framework for this effect in
outsourcing contracts must be a feasible suggestion for coming works.

References

Banker, R. D., and Kemerer, C. F. Scale Economics in New Software Development. IEE Transaction
on Software Engineering, 15, No. 10, 1199 – 1205, 1989.
Boehm, B. W. Software Engineering Economics. Englewood Cliffs, NJ: Prentice Hall. 1981.
Boehm, B. W., Abts, C., Brown, A. W., Chulani, S., Clark, B. K., Horowitz, E., Madachy, R., Reifer,
D. and Steece, B. Software Cost Estimation with COCOMO II. Upper Saddle River, NJ: Prentice Hall.
2000.
Bustinza, O.F., Arias-Aranda Gutierrez-Gutierrez D., L. Outsourcing, competitive capabilities and
performance: an empirical study in service firms. International Journal of Production Economics 126 (2):
276-288, 2010.
Cachon, G., and Haker, P. Competition and Outsourcing with Scale Economies. Management Science
48(10): 1314 – 1333. 2002.
Debabrata Dey, Ming Fan, Conglei Zhang. Design and Analysis of Contracts for Software
Outsourcing. Information Systems Research 21(1):93-114. 2010.
Deci, E. L. Effects of externally mediated rewards on intrinsic motivation. Journal of Personality and
Social Psychology, 1971, 18, 105-115.
Eisenhardt, K. M. Agency theory: An assessment and review. Academy of Management Review, 14(1),
57. 1989.
Frey, Bruno S., and Reto Jegen. Motivation Crowding Theory. Journal of Economic Surveys, 15: 589-
611. 2001.
Gneezy, U. and A. Rustichini. A fine is a price, Journal of Legal Studies, 29, 1-17. 2000.
Gneezy, Uri.. The W Effect of Incentives. Mimeo. The University of Chicago Graduate School of
Business. 2003.

3038
 XLVI SIMPÓSIO BRASILEIRO DE PESQUISA OPERACIONAL
Pesquisa Operacional na Gestão da Segurança Pública
16 a 19
Setembro de 2014
Salvador/BA

Greiff, Matthias and Egbert, Henrik and Xhangolli, Kreshnik. Pay What You Want – But Pay
Enough! Information Asymmetries and PWYW Pricing. Joint Discussion Paper Series in Economics by
the Universities of Aachen. 2013.
Jensen, Michael C. and Meckling, William H. Theory of the Firm: Managerial Behavior, Agency Costs
and Ownership Structure. Journal of Financial Economics (JFE), Vol. 3, No. 4: 305 - 360,1976.
Keil, P. Principal Agent Theory and its Application to Analyze Outsourcing of Software Development.
Proc. of the Int. Workshop on Economics-Driven Software Engineering Research (EDSER), St. Louis,
USA, IEEE Computer Society. May 2005.
Lacity, M.C., Willcocks, L.P. An empirical investigation of information technology sourcing practices:
lessons from experience. MIS Quarterly 22 (3), 363–408, 1998.
Lacity, Mary C., Khan S. A., Willcocks, L. P. A review of the IT outsourcing literature: Insights for
practice Journal of Strategic Information Systems 18 (2009) 130–146
Londeix, B. Cost Estimation for Software Development. Wokingham: Addison-Wesley. 1987.
Mookherjee D. and Ivan Png. Optimal Auditing, Insurance, and Redistribution.The Quarterly Journal
of Economics, vol. 104(2), pp. 399-415, 1989.
Nosenzo, Daniele; Offerman, Theo, Sefton, Martin and van der Veen, Ailko. Discretionary Sanctions
and Rewards in the Repeated Inspection Game. CeDEx Discussion Paper No. 2012-10.
Singer, Marcos and Donoso, Patricio. Contracting contractors. Journal of Business Research, Vol. 64,
Issue 3,Pp. 338-343, 2011.
Sommerville, Ian. Software engineering 9th Ed. Pearson Education, Inc. 2011.
Watjatrakul, B. Determinants of IS sourcing decisions: a comparative study of transaction cost theory
versus the resource-based view. Journal of Strategic Information Systems 14 (4), 389–415, 2005.
Willcocks, L., Feeny, D. IT outsourcing and core IS Capabilities: challenges at lessons at DuPont.
Information Systems Management 23 (1), 49–56, 2006.
Whang, S. Contracting for Software Development, Management. Science, 38, 2, 307-324. 1992.

3039